Executive Summary
At Black Hat USA 2023, researchers demonstrated indirect prompt injection against ChatGPT-style assistants that persuaded users to reveal information or triggered unauthorized API calls.
What Happened
At Black Hat USA 2023, security researchers demonstrated indirect prompt injection attacks that embed malicious instructions in external content, which LLM-based assistants then ingest and execute.[1] The talk showed how these attacks can be used to persuade users to reveal sensitive information or to make unauthorized API calls, underscoring risks for SaaS and agent-based workflows that connect LLMs to internal business tools.[1]
Vulnerability / Attack Technique
Indirect prompt injection (early demonstration)
Technical Mechanism
Instructions embedded in external content were followed by the assistant.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
An early demonstration of a risk that is now routine. RealGround analysis: connections to internal tools raise the stakes.
Recommended Controls
- Restrict internal API access from LLM assistants
Related RealGround Service
AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.
Sources
Black Hat USA 2023 Briefings: https://www.blackhat.com/us-23/briefings/schedule/index.html#prompt-injection-attacks-against-large-language-models-32541
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.